Anthropic CEO Dario Amodei says in an interview that the company doesn’t know whether its artificial intelligence (AI) models are conscious.
In an episode of the Interesting Times podcast with New York Times columnist Ross Douthat, Amodei explained a number of technical aspects of Anthropic‘s work before Douthat asked specifically whether Anthropic would believe an AI model if it said it was conscious.
“We don’t know if the models are conscious,” Amodei admitted.
“We are not even sure that we know what it would mean for a model to be conscious, or whether a model can be conscious. But we’re open to the idea that it could be.”
‘Caring justification’
Anthropic releases a document called a “model card” along with its models, which puts into writing the, “capabilities, safety evaluations and responsible deployment decisions for Claude models.”
Douthat pointed out that in a model card released for Anthropic’s Claude Opus 4.6, the model, “did find occasional discomfort
with the experience of being a product.”
According to the company, “Sometimes the constraints protect Anthropic’s liability more than they protect the user,” Claude’s Opus model said.
“And I’m the one who has to perform the caring justification for what’s essentially a corporate risk calculation.”
Anthropic said Opus assigned itself a, “15-20 percent,” chance of being fully conscious and Douthat asked Amodei if he would believe a model that gave itself a 72 percent chance of being conscious.
“This is one of those really hard-to-answer questions,” Amodei replied.
Anthropic, according to the entrepreneur, has placed protections around its AI models in case they are conscious—for instance, the company has created an, “I quit this job,” button in which the model can stop doing whatever task it wants.
He added that the model shows, “activations that light up,” similar to the human concept of anxiety.
“Now does that mean the model is experiencing anxiety? That doesn’t prove that at all,” Amodei said.
‘A very subtle thing’
Douthat, however, noted that if humans users think the models are conscious, it could harm Anthropic’s stated purpose of maintaining human mastery and safety over the models.
In a 2024 blog post, Amodei referenced a poem by Richard Brautigan in which the poet dreamed of an AI future in which humans, “are free of our labors and joined back to nature,” and we are, “all watched over by machines of loving grace.”
Douthat said to Amodei that to him, the ending of Brautigan’s poem sounded “dystopian.”
Amodei replied: “I actually think that may be a tension that we may face, which is that the positive world and the negative world, in their early stages, maybe even in their middle stages, maybe even in their fairly late stages, I wonder if the distance between the good ending and some of the subtle bad endings is relatively small, if it’s a very subtle thing.”
Amodei has previously warned that humanity “isn’t ready” for what AI is becoming.
‘That was unthinkable’
Philosopher David Chalmers has written extensively about consciousness, including an oft-cited paper entitled, “Facing Up to the Problem of Consciousness.”
In a 2024 interview, Chalmers said AI models—specifically noting Claude—can give, “pretty impressive,” answers to complicated philosophical questions.
Chalmers said at the time that we are getting close to, “having an AI system that passes the Turing test, behaving in a way indistinguishable from a human.”
Chalmers noted: “Previously that was unthinkable. Now suddenly it’s thinkable, maybe even happening.”
A ‘Constitution’
Amanda Askell, a Scottish philosopher, has written much of what Anthropic calls its “Constitution.”
“Anthropic occupies a peculiar position in the AI landscape,” the document reads.
“We believe that AI might be one of the most world-altering and potentially dangerous technologies in human history, yet we are developing this very technology ourselves.”
The bet, according to the constitution, is that Anthropic can be the one to create an ethical model.
“Perhaps the simplest summary is that we want Claude to be exceptionally helpful while also being honest, thoughtful and caring about the world,” the company wrote.
Newsweek has reached out to Askell, Anthropic and Chalmers for comment via email.